The dominance of the AI theme over stock-market performance, corporate profitability, economic growth and investor attention has grown so extreme that some very large structural questions now seem urgent and compulsory. Goldman Sachs came into the week with a comprehensive breakdown of this AI dominance across several fronts. On the narrowness of the S & P 500’s advance so far this year, the firm found “the TMT names within the S & P — comprising both the tech sector and communication services, plus Amazon and Tesla — account for 87% of the year-to-date rally.” Those same segments of the market that generated 87% of the S & P’s upside represent about 54% of the S & P 500’s weight. Framed this way, the market has not been “ignoring” or “shrugging off” the potential hazards of Middle East conflict, higher oil prices and rising bond yields. The parts of the market for which those things matter – the median consumer cyclical stock, which is down 14% from its high – are noticing such pressures. Wall Street’s bulls insist this divergence is logical and healthy because it reflects the evolving earnings outlook. True enough — by far, most of the upside to projected earnings is coming from the companies whose shares are pacing the rally. Goldman here plots the earnings-revisions trend for 2027 among AI-infrastructure plays, energy companies, the overall S & P 500 and the rest of the S & P outside of AI and energy. That final category has seen no lift at all to its profit outlook for next year since the start of 2026. Even if the fundamentals are pointing to a profit bonanza, share prices can overshoot. Hedge funds as a group are running with 20% of their net market exposure in semiconductor stocks, Jefferies’ trading desk pointed out on Monday. The high-momentum long-short basket – a strategy of owning the most persistent outperformers and shorting the deepest laggards – has only gained 20% or more over three months 11 times before today. The current move has far exceeded the average such surge, implying room to give back plenty of ground before even the underlying trend would be threatened. This Goldman chart ranks today’s episode against the average of those prior extreme upside moves. Beyond whether stocks are correct to capitalize massive anticipated profits in this way, there’s the question of whether this will inevitably come along with perilous imbalances in the real economy. AI capex next year is pegged to reach $1 trillion, or about 3% of U.S. GDP. While the trade-accounting impact on reported GDP might not fit to exactly that number (much of the AI hardware is imported and counts against domestic output), it’s a useful scale. At its peak, investment in railroads in the 19th century got to 5%-6% by many estimates and resulted in a bubble. Nvidia caution The size of this commitment is forcing even those who are betting heavily on a durable AI capex cycle to interrogate some of the underlying assumptions. Gabelli Funds portfolio manager John Belton details a potential cautionary case for Nvidia , the largest holding in the growth funds he oversees which reports Wednesday after the bell: “Roughly half of NVIDIA’s business comes from about five companies. Those five companies have collectively seen their free cash flow generation move close to zero this year. So, the negative case for the stock is essentially that for half of NVIDIA’s business to continue growing, those customers either need to become free-cash-flow negative or they need to see accelerating free cash flow generation. That’s the more complicated dynamic for NVIDIA at this stage, and I think it’s part of the reason the stock has lagged some of the other semiconductor names recently.” Is the market’s recent preference for the lower-value-added groups in tech (memory makers, Intel, CPU providers) a sustainable shift or simply reflecting a fleeting ability by these companies to exploit pockets of scarcity to impose a tax on the infrastructure builders? The voracious demand for capital among the builders of AI-computing capacity could also threaten to strain the ability of investors to provide it at a palatable price. The big tech platform companies have had no trouble finding buyers for their debt, but the issuers are increasingly ranging globally to place it. Coming at a time of war and large fiscal deficits globally, and with inflationary pressures building from scarcity across multiple industrial commodities, rates across the world are rising toward the top end of multi-year ranges. I’m not a real believer in “bond vigilantes,” investors collectively punishing governments for profligacy through higher rates. But the bond market can act as the neighbors who call the police when the party gets too rowdy. Sometimes the cops just tell the revelers to turn the music down a bit, other times they break up the bash. Shadow supply Then there’s the shadow supply of equity from the massive IPOs lining up to hit the markets. If SpaceX truly comes at the rumored $1.75 trillion market value, with OpenAI and Anthropic behind it at or near $1 trillion each, it would equate to roughly 7% of the S & P 500’s total value. Not that they would enter the index soon or at those levels. The index only weights companies by the value of their publicly floated shares, which could be small to start. But these deals would be sure to make the mega-cap tier of the market more volatile and fundamentally expensive once they arrive. The market has been happy to ride the capex-over-consumer trend, implicitly endorsing the notion that companies can continue to profitably build new capacity to earn better returns. This, even as corporate profits as a share of GDP are already at a record. These big-picture questions hover over a market also poised to work through some more prosaic, near-term issues. Would a reopening of the Strait of Hormuz prompt a pendulum swing away from tech and back to the rest of the tape, and what might that mean for the big-cap benchmarks? Are more acute vulnerabilities being exposed by the rise in yields, and will elevated inflation expectations create an undertow to equity valuations? And can the extreme technical divergences and multiple overbought conditions be ameliorated without much pain, as suggested by Monday’s benign rotational action in the face of a stiff drop in semis and other momentum stocks?
AI stocks helped the bull power through multiple threats. But now is this market too out of balance?
May 19, 2026
